A Reputation-Based Trust Model for Agent Societies

Abstract

Trust and security issues are prevalent in agent societies, where agents are autonomously owned and operated in a networked environment. Nowadays, trust and reputation management is a promising approach to manage them. However, many reputation models su.ered from a major drawback – there is no mechanism to discourage agents from lying information when making a recommendation. Although some works do take into account of this issue, they usually do not penalize an agent for making poor referrals. Worse, some systems actually judge an agent’s referral reputation based on its service reputation. In situations where this is unacceptable, we need to have a mechanism where agents are not only discouraged from making poor referrals, but are also penalized when doing so. Towards this, we propose a reputation-based trust model that considers an agent’s referral reputation as a separate entity within the broader sense of an agent’s reputation. Our objective is not to replace any existing reputation mechanisms, but rather to complement and extend them.